A Three-Component Model of the Control Error in Manual Tracking of Continuous Random Signals

Abstract

The performance of human operators acting within closed-loop control systems is investigated in a classic tracking task. The dependence of the control error (tracking error) on the parameters display gain, k(display), and input signal frequency bandwidth, f(g), which alter task difficulty and presumably the control delay, is studied with the aim of functionally specifying it via a model. The human operator as an element of a cascaded human-machine control system (e.g., car driving or piloting an airplane) codetermines the overall system performance. Control performance of humans in continuous tracking has been described in earlier studies. Using a handheld joystick, 10 participants tracked continuous random input signals. The parameters f(g) and k(display) were altered between experiments. Increased task difficulty promoted lengthened control delay and, consequently, increased control error.Tracking performance degraded profoundly with target deflection components above 1 Hz, confirming earlier reports. The control error is composed of a delay-induced component, a demand-based component, and a novel component: a human tracking limit. Accordingly, a new model that allows concepts of the observed control error to be split into these three components is suggested. To achieve optimal performance in control systems that include a human operator (e.g., vehicles, remote controlled rovers, crane control), (a) tasks should be kept as simple as possible to achieve shortest control delays, and (b) task components requiring higher-frequency (> 1 Hz) tracking actions should be avoided or automated by technical systems.